HAL Id: hal-02117441
https://hal.archives-ouvertes.fr/hal-02117441
Submitted on 27 Nov 2020
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Multi-faceted particle pumps drive carbon sequestration in the ocean
Philip W. Boyd, Hervé Claustre, Marina Lévy, David Siegel, Thomas Weber
To cite this version:
Philip W. Boyd, Hervé Claustre, Marina Lévy, David Siegel, Thomas Weber. Multi-faceted particle pumps drive carbon sequestration in the ocean. Nature, Nature Publishing Group, 2019, 568 (7752), pp.327-335. �10.1038/s41586-019-1098-2�. �hal-02117441�
1
Multi-faceted particle pumps drive carbon sequestration in the ocean 1
2 3
Revised for Nature 10 January 2019 4
5 6
Philip W. Boyd1 Hervé Claustre2, Marina Levy3, David A. Siegel4, Thomas Weber5 7
8
1Institute for Marine and Antarctic Studies, University of Tasmania, Hobart, Tasmania, 9
Australia 10
2 Sorbonne Université & CNRS, Laboratoire d'Océanographie de Villefranche-sur-mer 11
(LOV), 06230 Villefranche-sur-Mer, France.
12
3Sorbonne Université, LOCEAN-IPSL, CNRS/IRD/MNHN, 4 Place Jussieu, 75252 Paris 13
CEDEX 05, France.
14
4Department of Geography & Earth Research Institute, University of California, Santa 15
Barbara, Santa Barbara, CA, 93106, USA, 16
5Department of Earth and Environmental Sciences, University of Rochester, Rochester, NY 17
14627 18
19
Orchid # 20
Philip Boyd http://orcid.org/0000-0001-7850-1911 21
Hervé Claustre 0000-0001-6243-0258 22
Marina Levy 0000-0003-2961-608X 23
David Siegel https://orcid.org/0000-0003-1674-3055 24
Thomas Weber 0000-0002-4445-6742 25
26 27 28 29 30 31 32
2 33
The ocean’s ability to sequester carbon out of contact with the atmosphere exerts an 34
important control on global climate. The biological pump drives carbon storage in the 35
deep ocean and is thought to function via gravitational settling of organic particles from 36
surface waters. However, the settling flux alone is often insufficient to balance 37
mesopelagic carbon budgets or meet the demands of subsurface biota. Here, we review 38
additional biological and physical mechanisms that inject suspended and sinking 39
particles to depth. Together, these “particle injection pumps” likely sequester as much 40
carbon as the gravitational pump, closing carbon budgets and motivating further 41
investigation of their environmental controls.
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
3 58
Introduction 59
Open ocean waters store (sequester) carbon out of contact with the atmosphere on decadal to 60
millennial timescales, exerting a major control on global climate by regulating atmospheric 61
carbon dioxide partial pressure (pCO2)1. The magnitude of ocean carbon storage is governed 62
by two well-established mechanisms that maintain a surface-to-deep ocean gradient of 63
dissolved inorganic carbon (DIC) – the biological and the solubility pumps2,3. The solubility 64
pump delivers cold, dense, DIC-rich waters to depth mostly at high latitudes, whereas the 65
biological pump globally exports particulate organic carbon (POC) from surface waters. POC 66
export is largely attributed to the gravitational settling of a subset of the particle assemblage1,4 67
– a process we refer to as the “biological gravitational pump” (BGP).
68
The BGP is the key link between upper ocean photosynthetic carbon fixation, the sustenance 69
of mid-water biota, and carbon storage in the oceans’ interior4,5, andis thought to account for 70
~90% of the vertical DIC gradient, while the solubility pump explains the remainder1. In the 71
absence of the BGP, models predict atmospheric pCO2 would be higher by nearly twofold6. 72
Contemporary and paleoceanographic observations both reveal that carbon sequestration by 73
the BGP is affected by environmental changes in light, temperature, stratification and nutrient 74
availability7,8, and can itself drive dramatic climate shifts such as glacial-interglacial cycles8. 75
Future climate projections suggest that the functioning of the BGP will be altered by ocean 76
global change7,9, potentially feeding back on anthropogenic climate warming10. As a 77
consequence, quantification of its functioning requires a reliable baseline of accurate 78
measurements.
79
The underlying principles of the BGP are long established11: organic particles are continually 80
produced and recycled in sunlit surface waters, and a small fraction of these settle into the 81
4
oceans’ interior. The strength of the BGP is often quantified as the rate of particle “export”
82
from the euphotic zone, the surface mixed layer, or across an arbitrary horizon at 100m12. As 83
they sink, particles undergo myriad transformations, which lead to pronounced vertical 84
attenuation of the particle flux that is often described as a power law relationship, referred to 85
as the “Martin Curve”13. The efficiency of the BGP is defined here as the time that exported 86
carbon is kept sequestered from the atmosphere within the ocean’s interior. It is driven by the 87
depth scale of flux attenuation and pathways of ocean circulation that carry remineralized 88
carbon dioxide back to the surface14. Carbon is sequestered for timescales longer than a year 89
by particles that penetrate the permanent pycnocline (beneath the wintertime mixed layer) 90
and up to centuries by those that reach deep water masses (generally >1000m). Together, the 91
strength and efficiency of the BGP determine the total quantity of carbon sequestered 92
biologically in the ocean interior.
93
Recently, analyses of global and regional ocean carbon budgets have identified conspicuous 94
imbalances (i.e., two to three-fold less storage) when BGP export fluxes are compared with 95
those derived from geochemical tracers15,16, highlighting the need to reassess the pathways 96
that contribute to carbon storage. Furthermore, rates of site-specific particle export appear to 97
be insufficient to meet the carbon demand of mid-water life (termed mesopelagic biota) by 98
two-to three-fold17-20, but in one study can be balanced using community respiration18. There 99
is considerable debate over the reasons for these carbon deficits, ranging from biases inherent 100
in observational technologies17,21 to the potential role of other carbon (dissolved and/or 101
particulate) delivery mechanisms to deep waters16,22,23. Traditionally, the biogeochemical 102
functioning of the BGP has been evaluated from quasi one-dimensional (1D) observations of 103
particle flux (Box 1), and extrapolated using Earth System Models (ESMs, parameterised 104
with observations24-26) and/or remote-sensing observations26. This approach cannot capture 105
5
more complex mechanisms of carbon export that are highly variable in space and time (Box 106
1), potentially resulting in the reported carbon budget deficits.
107
Multiple lines of research have revealed the importance of additional export pathways, 108
physically (e.g. subduction) and/or biologically (e.g. large mesopelagic migrators) -mediated, 109
that inject particles to depth, termed here Particle Injection Pumps (PIPs)23,27-30. These 110
mechanisms can potentially export all particle classes to depth, and thus challenge the 111
conventional view of gravitational sinking as the dominant downward pathway for particles 112
into the oceans’ interior. The characteristics of PIPs fundamentally change our understanding 113
of biological carbon sequestration: first, PIPs can animate particle transport spatially into 114
three dimensions (3D), in contrast with the BGP where the vertical dimension is predominant 115
(1D); second, global estimates of PIP carbon fluxes are significant relative to those for the 116
BGP27,28, and third, these mechanisms cannot be readily quantified using the traditional 117
toolbox applied to investigate the BGP (Box 1). Overall, the PIPs will increase the strength 118
of the biological pump beyond estimates based on gravitational flux alone, and can change its 119
efficiency by altering the depth of carbon export.
120
The fate of exported carbon following its delivery to depth has also proven more complex 121
and heterogeneous than previously recognized. Particle flux attenuation is now known to vary 122
systematically in space14,31,32 and time33, suggesting the traditional empirical view13 must be 123
replaced by a mechanistic one that considers particle composition and architecture, microbial 124
metabolism, and transformation processes17. 125
Together, these developments stand to reshape our understanding of particle transport and 126
remineralisation in the oceans’ interior. Here, for open ocean systems we review: the 127
mechanisms, rates, and depths of particle injection by each PIP; the potential for each 128
mechanism to close observed deficits in ocean carbon budgets; and the corresponding 129
6
remineralisation depths of exported POC in the deep ocean. We finish by outlining future 130
research directions needed to synthesize these developments into a new mechanistic, four- 131
dimensional (4D) view of carbon export and sequestration. The review does not detail the 132
important role of dissolved organic carbon subduction22,23, nor cover the dark microbial 133
carbon pump34 or chemolithotrophy35 which have been reviewed elsewhere (S-Table 1).
134
135
Particle injection pump mechanisms 136
PIPs differ in their mechanisms, temporal-spatial scales (Fig. 1, Fig. 2a), and/or geographical 137
extent, but have common features: i) they can act on all particles from suspended to sinking 138
(Fig. 1); ii) they typically inject particles below the euphotic zone (i.e., the export depth for 139
the BGP), potentially reaching depths >1000m28-30 depending on the injection mechanism 140
(Fig. 1, Fig. 2b); iii) they occur concurrently with the BGP but cannot be measured with 141
techniques developed to quantify gravitational settling13,32 (Box 1); iv) their dynamic nature 142
(i.e., physical transport23,27,28 or patchiness of animal distributions30) means that the interplay 143
between their vertical and horizontal vectors and temporal scales varies significantly (Fig. 1).
144
Hence, a 4D sampling framework is required to constrain them (Box 1). The main 145
characteristics of each PIP are elucidated below.
146
Particle export driven by physical subduction includes several processes driving the vertical 147
transport of near-surface particles that act on different space/time scales: subduction caused 148
by mixed-layer shallowing (termed the mixed-layer pump29,36); subduction by large-scale 149
(100-1000 km) circulation (termed the large-scale subduction pump)23; and subduction by 150
mesoscale (10-100 km) to submesoscale (1-10 km) frontal circulation (termed the eddy- 151
subduction pump23,27,28).
152
7
Carbon export by the mixed-layer pump is driven by biological accumulation of particles 153
throughout the spring/summer growth season, which are then diluted to the depth of the 154
mixed layer during winter, and left in the oceans’ interior during early spring stratification 155
(Box 1). This pump operates on wide-ranging time-scales from days/weeks37 to seasons29,37, 156
predominantly in mid and high latitude regions characterised by strong seasonal variability in 157
mixed-layer depth (Fig. 2a). Although these concepts are long-established36, only recently 158
have they been scrutinised in detail using advances in optical profiling (BGC-Argo) floats 159
and satellite particle proxies to track particle accumulation rates in relation to changes in 160
surface mixed-layer depth (Box 1).
161
The large-scale subduction pump is a 3D advective mechanism directed from the seasonal 162
mixed-layer into the oceans’ interior, driven by Ekman pumping and horizontal circulation 163
across a sloping mixed-layer38. Subduction rates were first estimated for the North Atlantic39, 164
and then globally using data-assimilating models40. The wide-ranging subduction rates (1-100 165
m/year)39,40 are small relative to BGP particle settling rates11,12, but subduction occurs over 166
large regions of the global ocean boosting the magnitude of carbon delivery to depth.
167
The frontal-associated eddy-subduction pump subducts particle-rich surface waters on 168
timescales of days and across spatial scales of 1-10 km, driven by strong vertical circulation 169
associated with fronts and eddies27,28,41-44
. Gliders are now used to map 3D dynamic eddying 170
flow fields (Box 1), finding evidence for penetration of high particle stocks (co-located POC 171
and chlorophyll indicative of viable phytoplankton) from the spring bloom, conspicuous as 172
distinct filaments at 100-350 m depth at the eddy periphery28 (Box 1). Mapping revealed the 173
co-location of high POC filaments and negative vorticity to depths near the permanent 174
pycnocline28, and the mechanism is supported by high-resolution simulations in which eddy 175
subduction of particles is a recurring feature45-48. The strength of the eddy-subduction pump 176
8
is governed by the vigour and penetration of the vertical circulation, in conjunction with local 177
POC stocks over the frontal area27,49. Eddy subduction rates span 1-100 m d-1 (c.f. 20 to >
178
100 m d-1 for the BGP11,12) depending on the eddy or frontal structure. Modelling indicates 179
that these subducted particles are remineralised more rapidly (i.e., at relatively shallow depths) 180
relative to gravitationally-sinking particles27. 181
The concept for the ‘mesopelagic migrant pump’ is based on long-established observations of 182
diurnal vertical migration50 (Box 1). This pump extends the remineralisation scale by 183
injecting particles to greater depth before decomposition begins51,52, as determined by gut 184
retention time of migrating animals51-53 and the depth of their migration (typically ~400 m53).
185
The injected particles are zooplankton faecal pellets with sinking rates of 10-100’s m d-1 (ref.
186
51), faster than loosely-packed organic aggregates settling from the surface11,12, and will 187
penetrate deeper in the water column before remineralisation. This pump therefore influences 188
all important facets of the particle flux that govern carbon sequestration – total export rate, 189
depth of peak flux, and flux attenuation depth scale.
190
Diurnal vertical migration results in active subsurface transport and carbon sequestration, and 191
is usually reported for mesozooplankton and often included in BGP estimates51. However, 192
vertical migration by larger mesopelagic carnivorous organisms (from greater daytime depths 193
than mesozooplankton) are not sampled by conventional BGP approaches52,54. Targeted 194
studies (Box 1) have quantified this pump driven by large mesopelagic migrant carnivores in 195
the Pacific54, and other regions (S-Table 1). The underlying mechanism is upward migration 196
to graze mesozooplankton54 followed by rapid (hours) downward migration53, with 197
respiration (release of CO2), exudation, and defecation (release of POC/DOC)51,55 often 198
below the permanent pycnocline56, at depths up to 600m (Box 1).
199
9
Trawl surveys suggest that ~50% of mesopelagic organisms migrate, ranging regionally 200
between 20-90% depending on temperature, turbidity and oxygen concentrations54,56. The 201
carbon sequestration rate by this pathway is governed by the metabolic transfer efficiency of 202
migrators, and particles are injected at their residence depth, often at the upper boundary of 203
oxygen minimum zones where their respiration intensifies oxygen depletion53. 204
Active transport by vertically-migrating metazoans can also occur on longer timescales (Box 205
1). For example, in high latitude regions the winter hibernation of copepods (members of the 206
mesozooplankton) at depths between 600-1400m gives rise to a so-called ‘seasonal lipid 207
pump30’: during hibernation, they catabolise carbon-rich lipids accumulated during summer in 208
upper layers and thereby shunt carbon (but not nitrogen and phosphorus) below the 209
permanent pycnocline30. The strength of the seasonal lipid pump is governed by copepod 210
abundance, size and temperature, which together control their respiration rate and help 211
explain the existence of carbon flux hotspots (i.e. patchiness)30. 212
Another vertical export mechanism that operates on seasonal migration timescales is 213
mortality at depth of hibernating zooplankton particularly in high latitude regions57,58, 214
sequestering carbon to depths >500 m depth. Global extrapolation of seasonal lipid pump 215
fluxes, along with the over-wintering mortality flux is problematic due to difficulties in 216
sampling and generalizing across distinct regional mechanisms30 (S-Table 1).
217
218
The potential for double accounting 219
The export flux from the BGP is mediated by sinking particles, whereas PIPs can provide 220
additional pathways for all particle classes, from suspended to sinking, to exit the surface 221
ocean (Fig. 1). Thus, there is potential overlap between particles delivered from the surface 222
ocean to depth via the BGP and by injection from PIPs. Such overlap – termed here as 223
10
‘double-accounting’ – may occur where particles associated with the BGP and a PIP are 224
difficult to distinguish and hence could be attributed to more than one pump (Fig. 1). At 225
depth, transformations such as aggregation alter particle characteristics, including size and 226
sinking rate, and hence particles injected by the PIPs can join the sinking flux usually 227
attributed the BGP (Fig. 1). A further factor that introduces overlap between the BGP and 228
PIPs results from the inclusion, for historical reasons59, of one component of the mesopelagic 229
migration pump (diurnal migration by mesozooplankton) into the 1D sampling framework of 230
the BGP, while other components (e.g. patchier diurnal migration by larger mesopelagic 231
carnivores5) are not. Hence double-accounting can confound our understanding of the 232
relative importance of PIPs to ocean carbon storage.
233
Is it possible to tease apart these areas of overlap? Forty years study of the BGP has 234
uncovered a complex biogeochemical system with multiple drivers and distinguishing 235
characteristics11,60. This body of research helps to frame the differences and similarities 236
between particles delivered to depth by PIP’s and those settling via the BGP. Each PIP is 237
distinct with respect to its combination of injected particle type (suspended cells to faecal 238
pellets of large mesopelagic migrants), the timing and depth of injection (Fig. 2a-b), and 239
associated particle transformations (aggregation/disaggregation)11,12,61. Additionally, the 240
subsurface “fate” of particles (i.e. where they remineralize), which determines the longevity 241
of carbon sequestration, is driven by the complex interplay between these properties and 242
transformations12,60,61: Particle composition and architecture set their sinking speed, while 243
myriad processes that are biologically- (microbes/zooplankton) and physically-mediated 244
(fragmentation/ disaggregation)12,62-64 decompose and repackage them over depth (Fig. 1).
245
Therefore, particle fate provides another avenue to distinguish the contributions of PIPs from 246
the BGP.
247
11
To date, evidence on the subsurface fate of injected particles has been largely indirect27,28,49. 248
Surveys of eddy-subduction pumps suggest that injected particles may be remineralised at 249
depths <200 m, based on ammonium peaks49, time-series of biogeochemical gradients28, or 250
particle modelling studies27. In the NE Atlantic, reported high rates of particle 251
remineralisation (glider-based biogeochemical gradients) must be reconciled with concurrent 252
evidence of coincident, coherent chlorophyll plumes at depths >300 m indicative of 253
subducted viable phytoplankton28. This glider-based time-series reveals pronounced 254
patchiness28 suggesting that inference of the fate of injected particles even from state-of-the- 255
art observations is challenging.
256
Better constraining the contribution of each PIP to mesopelagic carbon budgets will require 257
characterisation of the injected particle assemblage and their transformations during 258
downwards transport12,65-68. Particle aggregation in PIPs may be driven by 259
convergence/subduction69-70 and/or differential sinking65,67, potentially leading to altered 260
modes of subsurface transport (Fig. 1). BGC-Argo profile observations allow quantification 261
of the size, type, seasonal succession, and penetration depths of particles injected by the 262
mixed-layer pump36 – properties which have the potential to differentiate them from fast- 263
sinking particles (i.e., BGP) whose distinctive ‘spiky’ bio-optical signature is readily detected 264
using multiple sensors71 (S-Figs. 2 and 3). Advances in bio-optics are already making cryptic 265
signatures associated with slow-sinking particles and zooplankton vertical migration less 266
opaque, lessening the possibility of double-accounting. Such double-accounting may be 267
avoided through the identification of unique characteristics of pumps including seasonality 268
(Fig. 2a), distinctive regional features30, or multi-variate oceanographic diagnostics72. 269
270
271
12 Carbon sequestration potential
272
The potential carbon sequestration by each PIP can be quantified as the product of their 273
carbon injection rate and their sequestration timescale, i.e. time until remineralised carbon is 274
returned to the surface (see Supplementary Methods). This timescale is determined both by 275
the injection depth of particles and their eventual fate, i.e. the degree to which they sink or 276
circulate through the ocean before remineralising to CO2. In general, deeper particle injection 277
and rapid sinking translates to longer carbon sequestration because the “passage time” from 278
the ocean interior to the surface increases with depth (Fig. 2b). Here, we assemble prior 279
estimates of carbon injection rate and depth (S-Table 1), along with new modelling 280
projections (Fig. 2), to estimate carbon sequestration by each PIP and assess their 281
significance relative to the BGP.
282
Some targeted studies provide concurrent estimates of carbon injection by individual PIPs 283
and the BGP27,28, whereas others30,54,57,58
facilitate comparison of regional-scale PIP fluxes 284
with independent estimates of the BGP. Both approaches reveal that PIPs each have the 285
potential to contribute significant rates of POC export. The reported upper bounds of global 286
PIP estimates summed together is 8.7 Pg C yr-1, which is comparable to the BGP export flux 287
(Table S1). This comprises 1.1-2.1 Pg C yr-1 for the large-scale/mesoscale physical pumps 288
(also includes DOC22,23), and 0.25-1.0, 0.9-3.6 and (-0.09) to 2.0 Pg C yr-1 from the lipid 289
seasonal, mesopelagic migration, and eddy-subduction pumps, respectively (Fig. 2c). Thus, 290
their cumulative contribution may be as much as ~40% of total particle export (i.e., 291
BGP+PIPs) suggesting considerable potential to resolve the imbalances reported for 292
mesopelagic carbon demand17, between nutrient and carbon export budgets15, and to lessen 293
the variability between model estimates of global carbon sequestration (S-Table 1).
294
13
We estimated the sequestration timescales for each PIP based on the “passage time” from the 295
injection depth to the surface in an observationally-constrained ocean circulation model14. 296
Particles injected at the depth of the wintertime mixed-layer by the large-scale physical 297
pumps (mixed-layer and subduction) result in sequestration for 25-100 years, assuming 298
subduction occurs before re-entrainment next winter. In turn, deeper injection by the eddy 299
subduction pump (up to 450 m), mesopelagic migration pump (up to 600 m), and seasonal 300
lipid pump (up to 1400 m) translates to sequestration timescales up to 150, 250, and 500 301
years respectively (Fig. 2b). These timescales will increase if it is assumed that sinking rather 302
than suspended particles are injected, which remineralise deeper than the injection horizon 303
(see Supplementary Methods).
304
Given the wide-ranging estimates of carbon injection rate (Fig. 2c) and depth (Fig. 2b) for 305
each PIP, oceanic carbon sequestration by these mechanisms cannot be estimated with 306
precision (Fig. 2d). However, choosing central values from the reported ranges of each 307
property allows a first order comparison between PIPs and the BGP. The mesopelagic 308
migration pump emerges as the most significant PIP, potentially storing ~60% as much 309
carbon as the BGP in the ocean interior if large, sinking particles (i.e. faecal pellets) are 310
injected. The C storage potential of the seasonal lipid, eddy-subduction and large subduction 311
pumps are ~20%, 10% and 5% of the BGP respectively, assuming each injects suspended 312
particles. The latter small net value is due to offsetting of subduction by strong obduction 313
(upward transport of water parcels) in the equatorial oceans39. Based on these central values 314
(Fig. 2d), it is likely that the reservoir of respired carbon in the ocean interior contributed by 315
the PIPs approaches that contributed by the BGP, and may therefore help to close global- 316
scale mesopelagic carbon budgets15,16. 317
318
14 Tracer constraints on the fate of exported carbon 319
Oceanic carbon sequestration by the BGP and wide-ranging biophysical mechanisms that 320
inject biogenic particles to depth depends critically on the fate of exported carbon (Fig. 2).
321
However, at present tracing the remineralisation of particles (regardless of their export 322
pathway) as they settle and circulate through the global ocean remains a logistical challenge, 323
due to the difficulties of deep-water particle sampling. Recently, new methods have used 3D 324
ocean data assimilation models to leverage geochemical “remineralisation tracers” including 325
oxygen and nutrients. These tracers integrate particle remineralisation signatures over long 326
timescales, and their global distributions are characterised by orders of magnitude more 327
observations than are available for particles16,31,73. Two distinct approaches have been applied.
328
The first diagnoses remineralisation rates directly from phosphate accumulation along 329
transport pathways in a circulation model, and reconstructs particulate flux profiles required 330
to explain the global distribution of remineralised phosphate31. The second assimilates 331
geochemical and satellite data into mechanistic biogeochemical models to optimise key 332
particle flux parameters, yielding mechanistic insights while leveraging the observations less 333
directly73. 334
Both approaches have yielded similar results and provide evidence for regional variability in 335
particle flux attenuation, with the flux attenuating slowly at high latitudes and quickly in 336
subtropical gyres, while the tropics lie between these two extremes (Fig. 3a). These 337
simulations reveal that carbon exported from high latitude and tropical surface waters is 338
sequestered longer in the oceans’ interior than carbon exported in the oligotrophic gyres 339
(Figure 3b), with important implications for feedbacks between the particle export and global 340
climate. Atmospheric pCO2 is likely more sensitive to past changes in high latitude export 341
than previously recognised8, and the future expansion of subtropical habitats9 may result in 342
less efficient (although not currently quantifiable) carbon sequestration in a warming world.
343
15
Regional variations in particle flux attenuation have largely been interpreted in terms of the 344
balance between decomposition and sinking rates32. A likely explanation for the diagnosed 345
latitudinal pattern is the temperature-dependent metabolism of heterotrophs responsible for 346
particle decomposition32,73, although variations in particle size and/or ballast are valid 347
alternatives73. There may also be a secondary effect of oxygen, with decomposition slowing 348
in anoxic zones73,74, and even hypoxic waters due to anaerobic microenvironment formation 349
in particles75. 350
To some degree, model-derived particle flux profiles may also reflect the relative magnitude 351
of different export pathways (PIPs and BGP), which vary in the injection depth and nature of 352
particles they supply, since geochemical tracers integrate the effects of all export mechanisms.
353
Deep injection by PIPs would result in slower flux attenuation over depth, whereas injection 354
of suspended particles that remineralise shallower in the water column would be diagnosed as 355
rapid flux attenuation. Predicting future changes in ocean carbon sequestration will require a 356
better understanding of the contribution of injection versus remineralisation processes to 357
sequestration efficiency (Fig. 3b), given the different environmental sensitivity of these 358
processes.
359
The need for prediction motivates development of new techniques to distinguish particle flux 360
associated with the BGP and each PIP. Particle stoichiometry (i.e., C:N:P) may be central to 361
identifying particular mechanisms that decouple their export. For example, diagnosing 362
oxygen consumption between 500-1500 m (depth of zooplankton hibernation) without 363
concomitant nutrient accumulation would point to carbon export by the seasonal lipid pump30. 364
Alternatively, diagnosing seasonal cycles of nutrient accumulation and oxygen consumption 365
rates would help distinguish remineralisation of particles exported by physical pumps versus 366
particle settling, which should exhibit distinct seasonality (Fig. 2a). This approach may soon 367
16
be possible given the burgeoning spatial/temporal resolution of tracer data provided by BGC- 368
Argo floats (S-Figure 1), and emerging float sensor technology (S-Table 2).
369
370
Extrapolation – towards a 4D view of particle export 371
Improving the accuracy of the initial estimates of the magnitude of carbon sequestration 372
presented in Figure 2d requires the development of a 4D picture of particle flux and storage 373
in the oceans’ interior. It is clear from our synthesis of PIP mechanisms that multiple scales, 374
from sub-mesoscale to basin, must be accommodated if PIPs are to be assembled, first 375
spatially and then temporally, into a complete 4D picture. Again, lessons on how to approach 376
such upscaling can be gleaned from BGP research which imprinted both spatial and seasonal 377
signatures (satellite remote-sensing/modelling)26 onto short-term (days-weeks) observations 378
taken at specific sites (Box 1). The timescales and lifetimes of features such as submesoscale 379
eddies/fronts or seasonal mesopelagic export signatures (Fig. 2a) must be characterized to 380
define the temporal footprint of each PIP and move towards a 4D viewpoint. This framework 381
must be linked to the seasonality of pelagic particle production to assess if there is distinctive 382
period for the subduction of significant stocks of these upper ocean particles (Fig. 2a). For 383
example, it is well-established that submesoscale dynamics are strongly seasonal, with 384
stronger and deeper penetration during winter than summer76. 385
Some published approaches towards extrapolating PIP’s globally, and to climatological time 386
scales, are outlined in S-Table 1. The identification of the specific drivers of each PIP 387
mechanism should help improve modelling and hence extrapolation. We advocate the utility 388
of explicitly incorporating the different PIP mechanisms into predictive, mechanistic models 389
as a means to extrapolate PIPs into 4D. In the case of the extrapolation of the submesoscale 390
eddy subduction PIP, increasing the model grid resolution to incorporate these features is 391
17
necessary and is now achievable in regional configurations77,78. In contrast, other physically- 392
mediated PIPs such as the large-scale subduction and mixed-layer pumps are already 393
represented in global models, and so their extrapolation requires the development of 394
diagnostics to enable the simulated POC/DOC distributions to be better evaluated against 395
observations23. At present, the biologically-mediated PIPs are not incorporated into state-of- 396
the-art biogeochemical models9,14,31,77,78. While simulating animal behaviour at the global 397
scale remains a grand challenge in ocean modelling, simple parameterisations have been 398
developed to predict the geochemical effect of the mesopelagic migrant pump6, which might 399
be further expanded to incorporate hibernation and therefore the seasonal lipid pump. It is 400
only very recently that diel vertical migration has been incorporated for the first time in a 401
global ocean general circulation model and used to estimate the associated flux of carbon at 402
the global scale (see Aumont et al. in S-Table 1). Although promising, this approach remains 403
challenging because it is based on a computationally-intensive, end-to-end ecosystem model 404
in which all trophic levels from phytoplankton to top predators interact.
405
406
Transforming our view of ocean carbon export 407
Our synthesis of physically- and biologically-mediated PIPs reveals that they are directly 408
transporting significant stocks of biogenic particles to depth, of a cumulative magnitude that 409
may be equivalent to the carbon storage of the BGP. The potential of PIPs to make a major 410
contribution to the ocean carbon budget must now be explored in more detail, commencing 411
with those PIPs that are most likely to contribute to carbon sequestration. Synthesising 412
estimates of particle export, injection depth, and circulation timescales reveals that the 413
mesopelagic migrant pump has the greatest potential to contribute to carbon sequestration, 414
followed by the seasonal lipid pump and the various physical pumps (Fig. 2d). In the case of 415
18
the seasonal lipid pump, its geographical realm of influence is already established30, whereas 416
less is known about the regional contributions of the mesopelagic migrant pump5. 417
For all PIPs, the most pressing research issue – needed to address double-accounting issues 418
and improve estimates of carbon sequestration – is to better understand the mechanisms of 419
particle transformations17,65-68 (Fig. 1) within a 4D framework. Specifically, the fate of 420
exported particles between their injection depth and the permanent pycnocline remains poorly 421
constrained. A first step will be improved particle characterisation, in particular the ability to 422
distinguish zooplankton from other particle types, and to aggregate Particle Size Distribution 423
(PSD) profiles through the development and application of new sensors (S-Table 2). Future 424
development of acoustic and imaging technologies79 must be deployed on a range of 425
platforms from ships (i.e., calibration) to an array of long-lived (i.e., years), geographically- 426
diverse BGC-Argo floats. These developments towards improving particle characterisation 427
will reduce the likelihood of double-accounting. Moreover, the alignment of BGC-Argo 428
deployments (Box 1) with the characteristic space and time scales of PIPs will enable better 429
quantification of the role of patchiness in driving observed local/regional hotspots in 430
biological PIPs30,54,56. In time, following the development and testing of a Coastal-Argo 431
platform, they can also be deployed to coastal and shelf seas to explore the role of PIPs in 432
these regions (S-Table 2).
433
The way forward in refining estimates of the contribution of PIPs in closing the ocean carbon 434
budget15-17 also requires leveraging advancements in ocean biogeochemical modelling.
435
Models are valuable testbeds to probe the sensitivity of carbon storage mechanisms, and 436
guide future observations. For example, model sensitivity analyses point to the pivotal role 437
of PSD in determining the fate of exported carbon31,73, but the processes that set the PSD of 438
exported particles and its evolution over depth remain crudely parameterized. Developing 439
robust models of particle transformations between multiple size classes, and incorporating 440
19
them into general circulation models, will allow us to trace the fate of particles injected by 441
different PIPS and dissect their contribution to carbon sequestration, while avoiding double- 442
accounting issues.
443
Inverse methods that can assimilate PSD fields from new BGC-Argo technologies80 will 444
allow models to “learn” from the data, further refining them to best reflect the real ocean.
445
Furthermore, downscaling of physical models is essential to simulate the locations of PIP 446
injections in support of observational programmes such as high resolution altimetry81, and the 447
integration of detailed particle transformations into submesoscale models82. 448
To transform the comprehension of particle export from one- to three- and eventually four- 449
dimensions, machine learning approaches83 will need to be employed, which can be trained to 450
predict unknown variables such as particle flux from better sampled variables. Approaches 451
like artificial neural networks84, will enable and enhance the upscaling of local/regional 452
datasets needed to provide more robust extrapolation85,86 to depth, regionally, and annually of 453
each PIP. This upscaling is essential to refine estimates of the contribution of each PIP to 454
carbon sequestration. BGC-Argo datasets will also eventually be combined with new satellite 455
products such as hyperspectrally-resolved ocean colour observations of biology processes87 456
and submesoscale characterisation of sea level using high-resolution altimetry81. 457
Satellite and water-column remote-sensing, along with targeted process studies, will yield 458
expansive datasets that can be assimilated into regional and global models of ever increasing 459
realism and resolution. Together, these approaches will lead towards a robust, four- 460
dimensional view of carbon sequestration by the ocean’s multi-faceted bio-physical particle 461
pumps.
462
463 464
20 Acknowledgements
465
The authors thank five anonymous reviewers for improving the manuscript. PWB was 466
primarily supported by the Australian Research Council through a Laureate (FL160100131), 467
and this research was also supported under Australian Research Council's Special Research 468
Initiative for Antarctic Gateway Partnership (Project ID SR140300001). HC acknowledges 469
the support of the European Research Council (remOcean project, grant agreement 246777) 470
and of the Climate Initiative of the BNP Paribas foundation (SOCLIM project). ML was 471
supported by CNES and by the ANR project SOBUMS (ANR-16-CE01-0014). DAS 472
acknowledges support from the National Aeronautics and Space Administration as part of the 473
EXport Processes in the global Ocean from RemoTe Sensing (EXPORTS) field campaign - 474
grant 80NSSC17K0692. TW was supported by NSF grant OCE-1635414. Co-authors, HC, 475
ML, DS and TW contributed equally to this Review.
476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498
21 References
499
1Sarmiento, J. L. & Gruber, N. (2006). Ocean Biogeochemical Dynamics. (Princeton 500
University Press, 2006).
501
2Volk, T., and M. Hoffert (1985). Ocean carbon pumps: Analysis of relative strengths and 502
efficiencies in ocean-driven atmospheric CO2 changes, in The Carbon Cycle and 503
Atmospheric CO2: Natural Variations Archean to Present, Geophys. Monogr. Ser., vol. 32, 504
edited by E. T. Sundquist and W. S. Broecker, pp. 99–110, AGU, Washington, D. C.
505
3McKinley GA, Pilcher DJ, Fay AR, Lindsay K, Long MC, Lovenduski NS. (2016).
506
Timescales for detection of trends in the ocean carbon sink. Nature 530:469–72 507
4Buesseler, K. O., Lamborg, C. H., Boyd, P. W., Lam, P. J., Trull, T. W., and co-authors 508
(2007). Revisiting carbon flux through the ocean's twilight zone through the ocean's twilight 509
zone. Science, 316, 567-570. doi: 10.1126/science.1137959.
510
5Irigoien, X. et al. (2014). Large mesopelagic fishes biomass and trophic efficiency in the 511
open ocean. Nat. Commun. 5,ncomms4271.
512
6Maier-Reimer E., U. Mikolajewicz, A. Winguth (1996). Future ocean uptake of CO2: 513
interaction between ocean circulation and biology. Climate Dynamics, 12, 711-721.
514
7Bopp, L., et al. (2013), Multiple stressors of ocean ecosystems in the 21st century:
515
Projections with CMIP5 models, Biogeosciences, 10, 6225–6245.
516
8Martinez-Garcia, A. et al. (2014). Iron fertilization of the Subantarctic Ocean during the last 517
ice age. Science 343, 1347–1350.
518
9Moore J.K. et al. (2018) Sustained climate warming drives declining marine biological 519
productivity. Science, 2018; 359 (6380): 1139 DOI: 10.1126/science.aao6379 520
10Bernardello R. et al. (2015) Response of the ocean natural carbon storage to projected 521
twenty-first-century climate changeJ of Climate DOI: 10.1175/JCLI-D-13-00343.1 522
11Boyd, P. W. & Trull, T. W. (2007). Understanding the export of biogenic particles in 523
oceanic waters: Is there consensus? Progress in Oceanography, 72(4), 276-312. doi:
524
10.1016/j.pocean.2006.10.007 525
12Buesseler, K.O., and Boyd, P.W. (2009). Shedding light on processes that control particle 526
export and flux attenuation in the twilight zone of the open ocean. Limnol. Oceanogr. 54, 527
1210–1232.doi:10.4319/lo.2009.54.4.1210 528
13Martin, J., G. Knauer, D. Karl, and W. Broenkow (1987), VERTEX: Carbon cycling in the 529
northeast Pacific, Deep Sea Res., Part A, 34, 267–285.
530
14DeVries, T., F. Primeau, and C. Deutsch (2012). The sequestration efficiency of the 531
biological pump, Geophysical Research Letters, 39, doi:10.1029/2012GL051963.
532
15Emerson, S. (2014) Annual net community production and the biological carbon flux in the 533
ocean. Global Biogeochemical Cycles 28, 1–12, doi:10.1002/2013GB004680 534
22
16Schlitzer, R. (2002) Carbon export fluxes in the Southern Ocean: results from inverse 535
modeling and comparison with satellite based estimates, Deep-Sea Research II, 49, 1623- 536
1644.
537
17Burd, A. B. et al. (2010). Assessing the apparent imbalance between geochemical and 538
biochemical indicators of meso- and bathypelagic biological activity: What the @$#! is 539
wrong with present calculations of carbon budgets? Deep-Sea Res. Part II Top. Stud.
540
Oceanogr. 57: 1557–1571. doi:10.1016/j.dsr2.2010.02.022 541
This paper reviewed the (lack of) progress on constraining mesopelagic carbon budgets, 542
and advocated new approaches to tackle this issue.
543
18Giering, S. L. et al. (2014). Reconciliation of the carbon budget in the ocean’s twilight 544
zone. Nature 507, 480–483. doi: 10.1038/nature13123 545
This paper presented one of the few balanced mesopelagic carbon budgets by assessing 546
community respiration versus carbon demand.
547
19Steinberg, D.K., B.A.S. Van Mooy, K.O. Buesseler, P. W. Boyd, T. Kobari, and D.M. Karl 548
(2008). Bacterial vs. zooplankton control of sinking particle flux in the ocean’s twilight zone.
549
Limnol. Oceanogr. 53: 1327–1338.
550
20Reinthaler, T. et al. (2006). Prokaryotic respiration and production in the meso- and 551
bathypelagic realm of the eastern and western North Atlantic basin. Limnol. Oceanogr. 51:
552
1262–1273.
553
21Boyd, P.W., McDonnell, A., Valdez, J. (2015) RESPIRE: An in situ particle interceptor to 554
conduct particle remineralization and microbial dynamics studies in the oceans' Twilight 555
Zone. Limnology and Oceanography-Methods Volume: 13s: 494-508.
556
22Hansell D.A., Carlson C.A., Repeta D.J., Schlitzer R., (2009). Dissolved organic matter in 557
the ocean. Oceanography 22:52–61.
558
23Lévy, M., Bopp, L., Karleskind, P., Resplandy, L., Ethé, C., & Pinsard, F. (2013). Physical 559
pathways for carbon transfers between the surface mixed layer and the ocean interior. Global 560
Biogeochemical Cycles, 27(4), 1001–1012. http://doi.org/10.1002/gbc.20092.
561
24Henson, S. A., Yool, A., & Sanders, R. (2015). Variability in efficiency of particulate 562
organic carbon export: A model study. Geophysical Res. Lett., 29, 33–45.
563
http://doi.org/doi:10.1002/2014GB004965 564
25Aumont, O., Van Hulten, M., Roy-Barman, M., Dutay, J.-C., Ethé, C., & Gehlen, M.
565
(2017). Variable reactivity of particulate organic matter in a global ocean biogeochemical 566
model. Biogeosciences, 14(9), 2321–2341. http://doi.org/10.5194/bg-14-2321-201713 567
26Siegel, D. A., K. O. Buesseler, S. C. Doney, S. F. Sailley, M. J. Behrenfeld, and P. W.
568
Boyd (2014), Global assessment of ocean carbon export by combining satellite observations 569
and food-web models, Global Biogeochem. Cycles, 28, 181–196, 570
doi:10.1002/2013GB004743.
571
27Stukel M.R., H. Song, R. Goericke, A.J. Miller (2017) The role of subduction and 572
gravitational sinking in particle export, carbon sequestration, and the remineralization length 573
23
scale in the California Current Ecosystem. Limnology and Oceanography, doi:
574
10.1002/lno.10636 575
28Omand, M.M. et al. (2015). Eddy-driven subduction exports particulate organic carbon 576
from the spring bloom. Science 348,222–225.doi:10.1126/science.1260062.
577
This paper quantified the Eddy Subduction Pump (ESP) using an array of gliders in the 578
North Atlantic during the spring bloom.
579
29Dall’Olmo G., J. Dingle, L. Polimene, R.J.W. Brewin and H.Claustre (2016). Substantial 580
energy input to the mesopelagic ecosystem from the seasonal mixed-layer pump. Nature 581
Geoscience, 9, 820-825 DOI: 10.1038/NGEO2818.
582
This paper quantified the Mixed Layer Pump (MLP) across large regions of the high 583
latitude ocean.
584
30Jónasdóttir S.H, Richardson K., Heath M.R. 2015. Seasonal copepod lipid pump promotes 585
carbon sequestration in the deep North Atlantic. PNAS 112:12122–26.
586
This paper provided the first detailed quantification of the seasonal lipid pump (SLP).
587
31Weber T. et al. (2016), Deep ocean nutrients imply large latitudinal variation in particle 588
transfer efficiency. PNAS, 113, 8606–8611.
589
32Marsay, C. M., R. J. Sanders, S. A. Henson, K. Pabortsava, E. P. Achterberg, and R. S.
590
Lampitt (2015), Attenuation of sinking particulate organic carbon flux through the 591
mesopelagic ocean, Proc. Natl. Acad. Sci. U.S.A, 112, 1089–1094.
592
32Giering, S. L. C., R. Sanders, A. P. Martin, S. A. Henson, J. S. Riley, C. M. Marsay, and D.
593
G. Johns (2017), Particle flux in the oceans: Challenging the steady state assumption, Global 594
Biogeochem. Cycles, 31, 159–171, doi: 10.1002/2016GB005424.
595
34Jiao N., et al. 2010. Microbial production of recalcitrant dissolved organic matter: long- 596
term carbon storage in the global ocean. Nat. Rev. Microbiol. 8:593–599.
597
35Swan, B.K. et al. (2011) Potential for Chemolithoautotrophy Among Ubiquitous Bacteria 598
Lineages in the Dark Ocean. Science, 333, 1296-1300.
599
36Bishop, J. K. B., M. H. Conte, P. H. Wiebe, M. R. Roman, and C. Langdon (1986), 600
Particulate matter production and consumption in deep mixed layers: Observations in a 601
warm-core ring, Deep Sea Res. Part A, 33, 1813–1841.
602
37Dall’Olmo, G., and K. A. Mork (2014), Carbon export by small particles in the Norwegian 603
Sea, Geophys. Res. Lett., 41, 2921–2927, doi:10.1002/2014GL059244.
604
38Cushman-Roisin, B. (1987). Subduction. Hawaii Univ, Dynamics of the Oceanic Surface 605
Mixed Layer P 181-196.
606
39Marshall, J., Nurser, A. & Williams, R. Inferring the subduction rate and period over the 607
North Atlantic. J. Phys. Oceanogr. 23, 1315–1329 (1993).
608
40Liu, L. L., Huang, R. X., 2012. (2012). The global subduction/obduction rates: Their 609
interannual and decadal variability. Journal of Climate, 25(4), 1096–1115.
610
http://doi.org/10.1175/2011JCLI4228.1 611
24 612
41Pollard R. T. & L. Regier (1990) Large variations in potential vorticity at small spatial 613
scales in the upper ocean. Nature 348, 227–229 doi:10.1038/348227a0.
614
42Nurser, A., & Zhang, J. (2000). Eddy-induced mixed layer shallowing and mixed 615
layer/thermocline exchange. Journal of Geophysical Research Ocean, 105(C9), 21851.
616
43Niewiadomska, K., Claustre, H., Prieur, L., & D’Ortenzio, F. (2008). Submesoscale 617
physical-biogeochemical coupling across the Ligurian Current (northwestern Mediterranean) 618
using a bio-optical glider. Limnol. Oceanogr, 53, 2210–2225.
619
44Estapa, M. L., D. A. Siegel, K. O. Buesseler, R. H. R. Stanley, M. W. Lomas, and N. B.
620
Nelson (2015), Decoupling of net community and export production on submesoscales in the 621
Sargasso Sea, Global Biogeochem. Cycles, 29, 1266–1282, doi:10.1002/2014GB004913 622
45Lévy, M, Klein, P. and A.-M. Treguer (2001). Impacts of sub-mesoscale physics on 623
phytoplankton production and subduction, J. Mar. Res., 59,535-565 doi:
624
10.1357/002224001762842181 625
46Nagai, T., Gruber, N., Frenzel, H., Lachkar, Z., McWilliams, J. C., & Plattner, G.-K.
626
(2015). Dominant role of eddies and filaments in the offshore transport of carbon and 627
nutrients in the California Current System. J. Geophys. Res. Ocean, 628
http://doi.org/10.1002/2015JC010889 629
47Karleskind, P., Lévy, M., & Memery, L. (2011). Modifications of mode water properties 630
by sub-mesoscales in a bio-physical model of the Northeast Atlantic. Ocean Modelling, 39, 631
47–60.
632
48Karleskind, P., Lévy, M., & Memery, L. (2011). Subduction of carbon, nitrogen, and 633
oxygen in the northeast Atlantic. Journal of Geophysical Research Ocean, 116(C2), C02025.
634
http://doi.org/10.1029/2010JC006446 635
49Stukel, M. R. et al. (2017). Mesoscale ocean fronts enhance carbon export due to 636
gravitational sinking and subduction. Proc. Natl. Acad. Sci. USA. 114: 1252–1257.
637
doi:10.1073/pnas.1609435114 638
This paper compared the magnitude of export fluxes from the biological pump and the 639
Eddy Subduction Pump (ESP).
640
50Vinogradov M.E. (1997) Some Problems of Vertical Distribution of Meso- and 641
Macroplankton in the Ocean. Advances in Marine Biology Volume 32, 1997, Pages 1-92.
642
https://doi.org/10.1016/S0065-2881(08)60015-2 643
51Steinberg D.K., and M.R. Landry (2017). Zooplankton and the ocean carbon cycle. Annu.
644
Rev. Mar. Sci. 2017. 9:413–44 645
52Bianchi D., Stock C., Galbraith E.D., Sarmiento J.L. (2013). Diel vertical migration:
646
ecological controls and impacts on the biological pump in a one-dimensional ocean model.
647
Glob. Biogeochem. Cycles 27:487–91 648
53Bianchi, D., Galbraith, E. D., Carozza, D. A., Mislan, K. A. S., & Stock, C. A. (2013).
649
Intensification of open-ocean oxygen depletion by vertically migrating animals. Nature 650
Geoscience, 6(7), 545.
651
25
This paper used global Acoustic Doppler Current Profiler observations to constrain the 652
Mesopelagic Migration Pump.
653
54Davison, P.C., Checkley Jr., D.M., Koslow, J.A., Barlow, J., (2013). Carbon export 654
mediated by mesopelagic fishes in the northeast Pacific Ocean. Progress in Oceanography 655
116, 14–30.
656
This paper used trawl surveys and metabolic modelling to assess the export fluxes 657
mediated by mesopelagic fishes.
658
55Childress, J. J., S. M. Taylor., G. M. Cailliet and M. H. Price (1980) Patterns of growth, 659
energy utilization and reproduction in some meso- and bathypelagic fishes off Southern 660
California. Marine Biology 61, 27-40 (1980) 661
56Klevjer T. A., X. Irigoien, A. Røstad, E. Fraile-Nuez, V. M. Benítez-Barrios & S.
662
Kaartvedt. (2016). Large scale patterns in vertical distribution and behaviour of mesopelagic 663
scattering layers. Scientific Reports, 6:19873, DOI: 10.1038/srep19873 664
57Bradford-Grieve JM, Nodder SD, Jillett JB, Currie K, Lassey KR. (2001). Potential 665
contribution that the copepod Neocalanus tonsus makes to downward carbon flux in the 666
Southern Ocean. J. Plankton Res. 23: 963– 75 667
58Kobari T, Steinberg DK, Ueda A, Tsuda A, Silver MW, Kitamura M. (2008). Impacts of 668
ontogenetically migrating copepods on downward carbon flux in the western subarctic Pacific 669
Ocean. Deep-Sea Res. II, 55:1648–60.
670
59Dam, H.G., C.A. Miller and S.H. Jonasdottir (1993) The trophic role of mesozooplankton 671
at 47N, 20W during the North Atlantic Bloom Experiment. Deep-Sea Res. II, 40, 197-212.
672
60Turner J.T. (2015). Zooplankton fecal pellets, marine snow, phytodetritus and the ocean’s 673
biological pump. Prog. Oceanogr. 130:205–48 674
61Bishop, J.K.B. (1989) Regional extremes in particulate matter composition and flux:
675
effects on the chemistry of the ocean interior. W.H. Berger, V.S. Smetacek, G. Wefer (Eds.), 676
Productivity of the ocean: present and past, Dahlem Konferenzen, John Wiley & Sons, New 677
York (1989), pp. 117-137 678
62McDonnell, A. M. P., P. W. Boyd, K. O. Buesseler (2015) Effects of sinking velocities and 679
microbial respiration rates on the attenuation of particulate carbon fluxes through the 680
mesopelagic zone. Global Biogeochemical Cycles. DOI: 10.1002/2014GB004935.
681
63Durkin, C.A., M.L. Estapa, and K.O. Buesseler (2015) Observations of carbon export by 682
small sinking particles in the upper mesopelagic. Marine Chemistry, 175, 72-81.
683
64Cavan EL, Trimmer M, Shelley F, Sanders R, (2017) Remineralization of particulate 684
organic carbon in an ocean oxygen minimum zone, Nature Communications, 8 Article 14847.
685
ISSN 2041-1723.
686
65Alldredge, A.L., Silver, M.W., (1988). Characteristics, dynamics and significance of 687
marine snow. Progress in Oceanography 20, 41–82.
688
26
66Jackson, G.A. (1990) A model of the formation of marine algal flocs by physical 689
coagulation processes. Deep Sea Research Part A. Oceanographic Research Papers 37 (8), 690
1197-1211, 1990.
691
67Kiørboe, T., (2001). Formation and fate of marine snow: small-scale processes with large- 692
scale implications. Scientia Marina 65 (Suppl. 2), 57–71.
693
68Iversen, M.H., Ploug, H., (2013). Temperature effects on carbon-specific respiration rate 694
and sinking velocity of diatom aggregates – potential implications for deep ocean export 695
processes. Biogeosciences 10, 4073–4085.
696
69Ohman M.D., R. Powell, M. Picheral and D.W. Jensen (2010) Mesozooplankton and 697
particulate matter responses to a deep-water frontal system in the southern California Current 698
System J. Plankton Res. 34, 815–827.
699
70D'Asaro, E. A. et al. (2018). Ocean convergence and the dispersion of flotsam. Proceedings 700
of the National Academy of Sciences, 30, 201718453–6.
701
http://doi.org/10.1073/pnas.1718453115 702
71Briggs, N., M. J. Perry, I. Cetinic, C. Lee, E. D’Asaro, A. M. Gray, and E. Rehm (2011), 703
High-resolution observations of aggregate flux during a sub-polar North Atlantic spring 704
bloom, Deep Sea Res. Part I, 58(10), 1031–1039.
705
72Stanley R.H.R., D.J. McGillicuddy Jr. Z. O. Sandwith, H. M. Pleskow (2017) 706
Submesoscale hotspots of productivity and respiration: Insights from high resolution oxygen 707
and fluorescence sections. Deep-Sea Research I, https://doi.org/10.1016/j.dsr.2017.10.005 708
73DeVries, T., and T. Weber (2017), The export and fate of organic matter in the ocean: New 709
constraints from combining satellite and oceanographic tracer observations, Global 710
Biogeochem. Cycles, 31, 535–555, doi:10.1002/2016GB005551.
711
74Cram, J.A., T. Weber, S.W. Leung, A.M.P. McDonnell, J._H. Liang, C. Deutsch (2018), 712
The role of particle size, ballast, temperature, and oxygen in the sinking flux to the deep sea.
713
Glob. Biogeo. Cyc. https://doi.org/10.1029/2017GB005710.
714
75Bianchi D., T.S.Weber, R. Kiko, C. Deusch (2018). Global niche of marine anaerobic 715
metabolisms expanded by particle microenvironments. NGEO in press.
716
https://doi.org/10.1038/s41561-018-0081-0. 717
76Callies, J., Ferrari, R., Klymak, J. M., & Gula, J. (2015). Seasonality in submesoscale 718
turbulence. Nature Communications, 6, 6862–9. http://doi.org/10.1038/ncomms7862 719
77 Lévy, M. et al. (2012). Large-scale impacts of submesoscale dynamics on phytoplankton:
720
Local and remote effects, 43-44(C), 77–93. http://doi.org/10.1016/j.ocemod.2011.12.003.
721
78Harrison, C. S., Long, M. C., Lovenduski, N. S., & Moore, J. K. (2018). Mesoscale effects 722
on carbon export: A global perspective. Global Biogeochemical Cycles, 32, 680–703.
723
https://doi.org/10.1002/2017GB005751 724
79Picheral, M., L. Guidi, L. Stemmann, D.M. Karl, G. Iddaoud Ghizlaine and G. Gorsky 725
(2010), The Underwater Vision Profiler 5: An advanced instrument for high spatial resolution 726
27
studies of particle size spectra and zooplankton, Limnol. Oceanogr. Methods, 8, 727
doi:10.4319/lom.2010.8.462.
728
80Johnson K. (2017). Biogeochemical sensors for autonomous, Lagrangian platforms:
729
Current status, future directions. Autonomous and Lagrangian Platforms and Sensors ALPS 730
II. https://alps-ocean.us/agenda/(last accessed 16 March 2017).
731
81Ubelmann, C. and L.-L. Fu, (2014) On the transition from profile altimeter to swath 732
altimeter for observing global ocean surface topography. J. Atmos. Oceanic Tech., 31, 560- 733
568.
734
82Resplandy, L. et al. (2012). How does dynamical spatial variability impact 234Th-derived 735
estimates of organic export? Deep Sea Res. I, 68(C), 24–45.
736
http://doi.org/10.1016/j.dsr.2012.05.015 737
83Castelvecchi, D. (2016). Can we open the black box of AI? Nature, 528, 20-23.
738
doi:10.1038/538020a.
739
84Sauzède, R. et al. (2016). A neural network-based method for merging ocean color and 740
Argo data to extend surface bio-optical properties to depth: Retrieval of the particulate 741
backscattering coefficient. Journal of Geophysical Research-Oceans, 121(4), 2552-2571.
742
doi:10.1002/2015jc011408.
743
85Landschützer, P., N. Gruber, D. C. E. Bakker, U. Schuster, S. Nakaoka, M. R. Payne, T.
744
Sasse, and J. Zeng (2013), A neural network-based estimate of the seasonal to inter-annual 745
variability of the Atlantic Ocean carbon sink, Biogeosciences, 10, 7793–7815, 746
doi:10.5194/bg-10-7793-2013.
747
86Landschützer, P., N. Gruber, D.C.E. Bakker, and U. Schuster (2014). Recent variability of 748
the global ocean carbon sink. Global Biogeochemical Cycles, 28(9), 927-949.
749
doi:10.1002/2014gb004853.
750
87Werdell, P.J., L. I. W. McKinna, E. Boss, S. G. Ackleson, S. E. Craig, W. W. Gregg, Z.
751
Lee, S. Maritorena, C. S. Roesler, C. S. Rousseaux, D. Stramski, J. M. Sullivan, M. S.
752
Twardowski, M. Tzortziou, and X. Zhang, (2018), An overview of approaches and challenges 753
for retrieving marine inherent optical properties from ocean color remote sensing. Prog.
754
Oceanogr. 160, 186–212.
755
88Boyd, P.W. et al. (2005). The evolution and termination of an iron-induced mesoscale 756
bloom in the northeast subarctic Pacific Ocean. Limnology and Oceanography 50, 1872–
757
1886.
758
89Ohman, M.D., J.-B. Romagnan (2016) Nonlinear effects of body size and optical 759
attenuation on Diel Vertical Migration by zooplankton. Limnol. Oceanogr. 61, 2016, 765–
760
770.
761
90Powell, J.R., and M.D. Ohman (2012) Use of glider-class acoustic Doppler profilers for 762
estimating zooplankton biomass J. Plankton. Res, 34, 563–568.
763
91Siegel, D. A., and W. G. Deuser. (1997). Trajectories of sinking particles in the Sargasso 764
Sea: Modeling of statistical funnels above deep-ocean sediment traps. Deep-Sea Res. Part I 765
Oceanogr. Res. Pap. 44: 1519–1541. doi:10.1016/S0967-0637(97)00028-9 766
28
92Siegel, D. A., E. Fields, and K. O. Buesseler (2008), A bottom-up view of the biological 767
pump: Modeling source funnels above ocean sediment traps, Deep Sea Res. Part I, 55(1), 768
108–127, doi:10.1016/j.dsr.2007.10.006.
769
93Llort, J., Langlais, C., Matear, R., Moreau, S., Lenton, A., Strutton, P.G., (2018).
770
Evaluating Southern Ocean carbon eddy-pump from biogeochemical Argo floats. Journal of 771
Geophysical Research: Oceans. https://doi.org/10.1002/2017JC012861 772
773 774 775
Figure 1 Interplay between particle characteristics, mode of export (BGP or PIP), delivery depth and larger scale ocean circulation for a range of pumps. In the upper panel, the box (top left) represents mixed-layer particle types, which either form large sinking particles (i.e., within the BGP, such as faecal pellets, marine snow) or are injected to depth (i.e., PIPs, suspended/ slow-settling heterogeneous particles and cells (i.e., including healthy, slow-sinking phytoplankton88)). The vertical yellow arrow signifies the BGP; black lines physically mediated PIPs; and purple lines biologically mediated PIPs. The delivery rates of particles to subsurface strata (in m d-1, ? denotes not known) are presented for each pump. Patchiness in the distribution of vertically-migrating animals (top right) plays a role in driving three-dimensional particle delivery to depth89,89, and is denoted by different fish or copepod stocks in the upper ocean. The box (middle left) presents different particle transformations central to the BGP12, but whose role is not known so far for PIPs. They include microbial solubilisation, aggregation (marine snow denoted by aggregation 1;
heterogeneous faecally-dominated aggregates (aggregation II) and/or dissaggregation18 to form/break down heterogeneous particles (hatched brown symbols). In the lower panel, depths in parentheses are the reported delivery depths, with the BGP (and some PIPs) exporting some particles to the sea floor. Blue curved arrows represent transport of subsurface material along downward-sloping isopycnals (white dashed lines). Major unknowns include whether physical transport by PIPs can cause particle aggregation (signified by ? in the middle panel below subduction pump, and also applicable for the mixed-layer pump) and hence alter their mode of injection towards gravitational settling (i.e., the BGP). Other unknowns include the potential ballasting role of small mineral particles such as aerosol dust for PIPs.